diff --git a/tensorflow/core/lib/core/threadpool.cc b/tensorflow/core/lib/core/threadpool.cc
index 59d77a37344..a2245bb28ec 100644
--- a/tensorflow/core/lib/core/threadpool.cc
+++ b/tensorflow/core/lib/core/threadpool.cc
@@ -123,15 +123,6 @@ void ThreadPool::ParallelFor(int64 total, int64 cost_per_unit,
   impl_->ParallelFor(total, cost_per_unit, std::move(fn));
 }
 
-void ThreadPool::ParallelForWithWorkerId(
-    int64 total, int64 cost_per_unit,
-    const std::function<void(int64, int64, int)>& fn) {
-  impl_->ParallelFor(total, cost_per_unit,
-                     [this, &fn](int64 start, int64 limit) {
-                       fn(start, limit, CurrentThreadId());
-                     });
-}
-
 int ThreadPool::NumThreads() const { return impl_->NumThreads(); }
 
 int ThreadPool::CurrentThreadId() const { return impl_->CurrentThreadId(); }
diff --git a/tensorflow/core/lib/core/threadpool.h b/tensorflow/core/lib/core/threadpool.h
index 4d791721b43..371339598d3 100644
--- a/tensorflow/core/lib/core/threadpool.h
+++ b/tensorflow/core/lib/core/threadpool.h
@@ -47,7 +47,7 @@ class ThreadPool {
   // Schedule fn() for execution in the pool of threads.
   void Schedule(std::function<void()> fn);
 
-  // ParallelFor shards the "total" units of work assuming each unit of work
+  // ParallelFor shards the "total" unit of work assuming each unit of work
   // having roughly "cost_per_unit" cost, in cycles. Each unit of work is
   // indexed 0, 1, ..., total - 1. Each shard contains 1 or more units of work
   // and the total cost of each shard is roughly the same.
@@ -60,15 +60,6 @@ class ThreadPool {
   void ParallelFor(int64 total, int64 cost_per_unit,
                    std::function<void(int64, int64)> fn);
 
-  // Shard the "total" units of work. For more details, see "ParallelFor".
-  //
-  // The function is passed a thread_id in the range [0, NumThreads()]. The
-  // functions can safely write to a partial result for their id, in a tensor of
-  // size (NumThreads(), ...).
-  void ParallelForWithWorkerId(
-      int64 total, int64 cost_per_unit,
-      const std::function<void(int64, int64, int)>& fn);
-
   // Returns the number of threads in the pool.
   int NumThreads() const;
 
diff --git a/tensorflow/core/lib/core/threadpool_test.cc b/tensorflow/core/lib/core/threadpool_test.cc
index cecdc343799..c7d8db51364 100644
--- a/tensorflow/core/lib/core/threadpool_test.cc
+++ b/tensorflow/core/lib/core/threadpool_test.cc
@@ -80,42 +80,6 @@ TEST(ThreadPool, ParallelFor) {
   }
 }
 
-TEST(ThreadPool, ParallelForWithWorkerId) {
-  // Make ParallelForWithWorkerId use as many threads as possible.
-  int64 kHugeCost = 1 << 30;
-  for (int num_threads = 1; num_threads < kNumThreads; num_threads++) {
-    fprintf(stderr, "Testing with %d threads\n", num_threads);
-    const int kWorkItems = 15;
-    bool work[kWorkItems];
-    ThreadPool pool(Env::Default(), "test", num_threads);
-    for (int i = 0; i < kWorkItems; i++) {
-      work[i] = false;
-    }
-    std::atomic<bool> threads_running[kNumThreads];
-    for (int i = 0; i < num_threads; i++) {
-      threads_running[i] = false;
-    }
-    pool.ParallelForWithWorkerId(
-        kWorkItems, kHugeCost,
-        [&threads_running, &work](int64 begin, int64 end, int64 id) {
-          // Store true for the current thread, and assert that another thread
-          // is not running with the same id.
-          ASSERT_FALSE(threads_running[id].exchange(true));
-          for (int64 i = begin; i < end; ++i) {
-            ASSERT_FALSE(work[i]);
-            work[i] = true;
-          }
-          ASSERT_TRUE(threads_running[id].exchange(false));
-        });
-    for (int i = 0; i < kWorkItems; i++) {
-      ASSERT_TRUE(work[i]);
-    }
-    for (int i = 0; i < num_threads; i++) {
-      ASSERT_FALSE(threads_running[i]);
-    }
-  }
-}
-
 static void BM_Sequential(int iters) {
   ThreadPool pool(Env::Default(), "test", kNumThreads);
   // Decrement count sequentially until 0.